光谱液滴分析技术在液体安检方面的应用研究

发布时间：2018-06-13 18:37

本文选题：液滴分析技术 + 近红外光谱　；参考：《南京信息工程大学》2017年硕士论文

【摘要】：随着当前国际环境和形势的变化,公民在公共场合的人身安全问题变得日益突出。目前各交通领域的安检设备主要都是基于X射线成像原理来获取乘客携带物的信息,而大部分安检通道使用开盖试喝和气味检测的方法检测液态携带物,因此对液态危险物的检测是安检领域所重点关心的技术。本文将光纤液滴分析方法、电容液滴分析方法和近红外光谱分析方法相结合,研究了各方法鉴别危险液体实施方案,并通过实验结果分析判定各方案的可行性。本文首先构建了光谱液滴分析系统,设计了光纤信号检测电路和电容信号检测电路,通过硬件电路和采集软件的结合,完成了液滴检测实验平台的搭建。同时使用光谱仪对液体样本的吸光度信息进行采集。使用实验平台做初步实验,对一些典型的液体样本进行检测分析,验证该系统的可用性。运用仿真技术对样本的光纤、电容、光谱信号进行融合,得到样本的三维指纹图,从中提取特征参数并检验这些参数的识别能力。基于采集到的光谱数据,研究了近红外光谱技术对易燃液体快速鉴别的方法。采集了典型易燃液体和非易燃液体样品的近红外吸收光谱图,对数据进行平滑处理和基线校正。在建立判别预测模型时,对液体样品的光谱数据进行主成分分析,选取了特征波长点,从而提取样本的特征参数。利用距离判别、贝叶斯判别、Fisher判别三种方法建立了判别预测模型,结果表明Fisher判别法既准确又快速。同时,针对这些样本的光纤、电容液滴指纹图,进行去噪处理和归一化,选用波形分析法提取特征参数,再使用极限学习机算法构建分类判别模型。验证了光纤电容液滴系统对危险液体鉴别的可行性。
[Abstract]:With the change of the current international environment and situation, the personal safety of citizens in public has become increasingly prominent. At present, security equipment in various traffic fields is mainly based on the principle of X-ray imaging to obtain information on passengers' carriers, and most security channels use open lid testing and odor detection methods to detect liquid carriers. Therefore, the detection of liquid hazardous substances is a key technology in the field of security inspection. In this paper, the optical fiber droplet analysis method, capacitive droplet analysis method and near infrared spectrum analysis method are combined to study the scheme of identifying dangerous liquid by each method, and the feasibility of each scheme is determined by the analysis of experimental results. In this paper, the spectral droplet analysis system is constructed, and the optical fiber signal detection circuit and the capacitance signal detection circuit are designed. Through the combination of the hardware circuit and the acquisition software, the experimental platform of the liquid drop detection is built. At the same time, the absorbance information of liquid sample was collected by spectrometer. Some typical liquid samples are tested and analyzed by using the experimental platform to verify the usability of the system. Using the simulation technology, the fiber, capacitance and spectral signals of the sample are fused, and the three-dimensional fingerprint of the sample is obtained, from which the characteristic parameters are extracted and the recognition ability of these parameters is tested. Based on the collected spectral data, the method of fast identification of flammable liquids by near infrared spectroscopy (NIR) was studied. The near infrared absorption spectra of typical flammable and non-flammable liquid samples were collected, and the data were smoothed and baseline corrected. When the discriminant prediction model is established, the spectral data of the liquid sample are analyzed by principal component analysis, and the characteristic wavelength points are selected to extract the characteristic parameters of the sample. The discriminant prediction model is established by using the three methods of distance discrimination and Bayesian discriminant / Fisher discriminant. The results show that Fisher discriminant is accurate and fast. At the same time, according to the fiber and capacitance drop fingerprint of these samples, the denoising and normalization are carried out, and the characteristic parameters are extracted by waveform analysis method, and then the classification and discrimination model is constructed by using the extreme learning machine algorithm. The feasibility of identifying dangerous liquids with fiber capacitor droplet system is verified.
【学位授予单位】：南京信息工程大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;D035.3

【参考文献】